Why do we happily let AI suggest a hotel, but tense up the moment it tries to book one for us?

This explores why travelers are comfortable letting AI suggest things but get uneasy when it starts doing things for them. The simple answer is "higher stakes, lower trust." The corpus points to something more specific. In McKinsey and Skift's survey of about a thousand travelers, more than 90% said they trusted AI travel information overall. That comfort dropped sharply for tasks like checking visa requirements compared with open-ended brainstorming, and the reason travelers gave was hallucination Do travelers trust AI differently for brainstorming versus visa tasks?. Brainstorming forgives mistakes because a bad suggestion just gets ignored. A wrong visa answer can leave you stuck at the airport.

The sharper finding comes from a small study of students handing tasks to a general-purpose AI agent. Stakes alone did not make trust drop. Two other things did: whether the action could be undone, and whether other people would see it. Sending an email triggered sudden demands for approval even when users rated the output as fine. High-stakes tasks that could still be corrected caused no such reaction What makes people distrust AI agents they delegate to?. Applied to travel, this explains the line between advisor and agent well. An advisor's output stays private and can always be revised. An agent that books a non-refundable fare or emails a hotel takes an action you can't take back, out in the world. Confidence seems to depend less on how important the trip is and more on whether a mistake can be undone before anyone else sees it.

This fits a wider pattern. Capgemini found enterprise trust in fully autonomous agents fell from 43% to 27% in a single year as deployments moved past pilots, while organizations still planned to use agents in human-supervised teams Why is enterprise trust in autonomous AI agents declining?. One argument says the risk to people grows steadily with the autonomy handed over, and it favors a governed range of autonomy levels over a simple choice between advisor and agent Does AI risk increase with the autonomy we give it?. On the consumer side, shoppers accept high-autonomy assistants when the benefit is obvious and urgent, such as when scarcity makes speed valuable, and not because they feel more in control What drives shoppers to accept autonomous AI assistants?. A last-seat flight is the kind of moment that could tip a traveler toward letting the agent act.

There is also a warning about how far stated confidence can be taken at face value. Across languages, users follow how confident the AI sounds rather than whether it is right, so overconfident errors get acted on consistently Do users worldwide trust confident AI outputs even when wrong?. Advisory tools leave a human between that misplaced confidence and the consequences. Autonomous agents remove that human. The industry is quietly building around this: AI agents are not removing online travel agencies. They are pushing them deeper into the booking chain as the merchant of record, the party that settles payment and handles refunds Are AI agents actually removing online travel agencies from bookings?. That may be how the gap actually closes. Instead of travelers learning to trust the agent itself, the bookings it makes become reversible through a familiar intermediary.

One limit: the corpus has only one direct travel-traveler survey, and the reversibility finding comes from a 20-person study outside travel. The link between the two is a strong hypothesis, not a measured result for travelers.